Research On Extracting Classification Rules Of Hypertension Based On C4.5Algorithm

Hypertension is chronic disease, it can impact on human health seriously, and its prevalence on a global scale tends to go up, but the rate of treatment of patients with hypertension and awareness of hypertension medical knowledge at a low level. At present, the prevention and treatment of hypertension has also attracted the attention of the World Health Organization and scholars in medicine and other fields on a worldwide scale. With the development of information management of the health care industry, more and more hypertension clinical data are stored in the medical database. The problems to be solved is how to find the valuable information from the large amount of historical clinical data contribute to disease diagnosis and prediction in the future. In order to solve the problem, many scholars in recent years applied the data mining technology to the field of medicine, medical data mining is currently one of the hottest research subjects.This article tries to use data mining algorithms to learn Pathogenic Law from the extensive clinical data of patients with hypertension, and to find the main factor which can influence the blood pressure, then extracting the classification rules of hypertension. Because of the mining task is to classify, firstly we learned and compared the typical classification algorithms and their application in medicine. According to the continuous characteristics of data and the advantages of the decision tree model, we choose the C4.5algorithm to extract classification rules from the hypertension data.Then, this article gets clinical data of hypertensive patients entered into database, which includes patient’s electronic medical records, data of laboratory tests and electronic prescription data. But these raw data have varying degrees of incompleteness, noise and inconsistencies, and then processed the raw data with data cleansing, data transformation and data reduction and the other pretreatment techniques, finally chose the attribute which is related with the mining theme and drew a unified view.After the detailed description of the working principle of the C4.5algorithm and the methods of assessment of classification results, this paper analyzed the processed hypertension data with the C4.5algorithm, and constructed a decision tree model, and then extracted the classification rules to understand easier, and test the accuracy of the classification results with the maintain method. The experimental results demonstrate the applicability of the C4.5algorithm for hypertension classification rules extraction.Finally, for the short of the selecting in property of C4.5algorithm, we make the introduction of the concept of "relevant degree", and made changes to the procedures of the C4.5algorithm, to amend the information gain of the selected attribute, and change the support degree to the decision attribute of this attribute. The improved decision tree model is more in line with the medical understanding, the rightness rate also be improved accordingly, proving the validity of our improved algorithm.This article extracted classification rules from hypertension data, aimed at finding the properties of laboratory examination of the impact of hypertension and its dangerous level, to support physicians or patients on hypertension prevention and diagnosis.